Struct LocalCommandExecutor 

pub struct LocalCommandExecutor<W> { /* private fields */ }

Local non-PTY, one-shot command executor backed by host process spawning.

This executor resolves CommandRequest::cwd through the Workspace, starts argv directly without implicit shell parsing, applies EnvPolicy, requires a timeout, captures stdout/stderr separately, and reports output truncation.

This executor is not an OS sandbox. It does not technically enforce SandboxPolicy, NetworkPolicy, network isolation, process-tree cleanup, or resource limits. Its capability report marks sandbox, network, and process cleanup as advisory, and resource limits as unsupported.

By default, command execution fails closed when the requested policy requires technical enforcement that this local process backend cannot provide. LocalCommandExecutor::with_advisory_policy and LocalCommandExecutor::with_policy_capability_mode can explicitly allow such execution to continue with advisory enforcement reports. That mode is intended for tests, local prototypes, and reference CLI dogfooding; production coding-agent hosts should inject a CommandExecutor backed by a container, VM, platform sandbox, or remote isolated worker that can enforce the requested policy.

Implementations

impl<W> LocalCommandExecutor<W>

pub fn new(workspace: W) -> Self

Constructs a local command executor for a workspace.

pub fn with_advisory_policy(self, allow: bool) -> Self

Allows policy requirements that cannot be technically enforced locally to be reported as advisory instead of failing closed.

Passing true does not enable sandboxing, network isolation, process-tree cleanup, or resource limits. It only permits execution to continue while recording the downgrade in the returned policy enforcement report.

pub fn with_policy_capability_mode(self, mode: PolicyCapabilityMode) -> Self

Sets how this executor handles policy/capability mismatches.

PolicyCapabilityMode::RequireEnforced is the conservative default. PolicyCapabilityMode::AllowAdvisory allows local execution to proceed without technical enforcement and requires reports to say so explicitly.

Trait Implementations

impl<W: Clone> Clone for LocalCommandExecutor<W>

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<W> CommandExecutor for LocalCommandExecutor<W>

where W: Workspace,

fn capabilities(&self) -> CommandExecutorCapabilities

Returns executor capabilities.

fn execute<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, request: CommandRequest, policy: &'life1 ExecutionPolicy, context: &'life2 RunContext, ) -> Pin<Box<dyn Future<Output = Result<CommandOutput, CommandError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Executes a command under a harness execution policy.

impl<W: Debug> Debug for LocalCommandExecutor<W>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.